ii2030 Challenge: AI for Earth Observation. Who are the users?

Discover data usage scenarios for open machine learning ready Earth observation repositories like Radiant MLHub.

In about one month, we will launch the inclusive innovation 2030 (ii2030) process, where key stakeholders will assemble to turn a systemic challenge into an opportunity for all of us. We will attempt to answer a critical question:

“How might we ensure open access to high-quality machine learning (ML) ready Earth observation (EO) data on a sustainable basis?”

The ii2030 challenge will support open repositories like Radiant MLHub, which hosts a collection of geospatial training datasets and ML models. To facilitate the development of advanced analytics and near real-time insights from geospatial data, data engineers, developers, and data scientists need to be empowered by ML tools and datasets. Non-commercial as much as commercial organizations can then realize a whole range of use cases generating impact and value across geographies and industries. Let’s take a closer look at some great uses in the following fields:

  • Agriculture
  • Climate Impact and Modeling
  • Ecosystem Health Monitoring

Data usage scenarios for open ML-ready EO repositories


Our previous blog post focused on agriculture, telling the story of Gedeon Jean, a machine learning research engineer in Rwanda. He used various training datasets on Radiant MLHub to build a model that farmers can use to predict crop yield, identify disease risks, and monitor crop life cycles. The end users for Gedeon’s application are farmers, allowing them to make data-driven decisions. However, other types of industries can also generate value from these analyses, e.g.:

  • Banks can offer tailored lending instruments, and
  • Insurance companies may further customize their solutions for crop failures.

A case in point is CropIn’s SmartRisk, a commercial digital platform funded partially by the Bill & Melinda Gates Foundation. They leverage ML and EO to forecast farmers’ crops. The crop yield estimates are then used to help farmers to get better loans and insurance policies. Farmers can also use this application to help them maximize crop yield. At the same time, lending and insurance companies can quantify risks based on forecasts, while government agencies can plan better by evaluating and monitoring risks in real-time.

Other types of data usage would be to predict crop types and land uses in an area of interest. Crop and land use classifications are essential for monitoring services related to the environment and economic activities. Governments can also use it to build applications that allow more frequent updates to monitor agrarian change trajectories.

Last year, the Western Cape Department of Agriculture in South Africa joined forces with us (and others), inviting the world’s data scientists to build models that can identify crop type classes for their province. They intend to use the final application to develop an operational agricultural monitoring system in a region that experienced a frightening ‘Day Zero’ water crisis in 2018 and is a leading agricultural exporter of South Africa.

Climate Impact and Modeling

Climate impact and modeling is another domain for those looking at extreme events like flood and wildfire mapping or estimating the strength of tropical windstorms. In these scenarios, data practitioners are looking to find patterns in remote sensing data that can forewarn when a disaster is more likely to occur. Alternatively, a research practitioner in academia may be interested in analyzing changing trends and generating projections for climate indicators like atmospheric temperature and precipitation.

Flood mapping has gained momentum in recent years due to a 25% increase in flood events worldwide. In this domain, accurate flood prediction maps are necessary for all sectors. For example, government boards can use it to maintain flood insurance rate regulations and communicate the risks to local communities. On the other hand, private companies like lenders can determine insurance requirements, whereas insurance agencies can use these maps to underwrite parametric flood policies to supplement traditional ones.

Cloud to Street, a company founded by a group of research scientists, is an example of a flood mapping platform for flood disaster responses and risk analysis. Practitioners can access the data via their website or Google Earth Engine to create comparable applications. In addition, Radiant MLHub also hosts flood-related training datasets for data practitioners who want to design flood maps, which include the Cloud to Street — Microsoft Floods and Clouds training dataset.

Data products in the realm of wildfire mapping include but are not limited to active fire detection and wildfire prediction based on live fuel moisture data (available on Radiant MLHub for the western United States). North American and European governments have taken the lead in developing wildfire monitoring maps. The Canadian Wildland Fire Information and the EU’s forest fire monitoring systems are good examples of government investments. But many other countries lack these monitoring systems. In this data usage scenario, open repositories become even more critical as a benchmark because they allow users to measure their applications’ performance.

Ecosystem Health Monitoring

Anthropogenic activities such as deforestation are significant contributors to climate change. Therefore, monitoring changes to human ecology is another critical data usage scenario. Research institutions can build applications to identify waterholes for wildlife, measure and monitor the natural water cycle, or build deforestation alert systems based on tree cover loss. The global nonprofit, World Resource Institute, has created various conservation-related monitoring systems that incorporate remote sensing as a critical element, many made with open EO data.

Practitioners can also build applications to detect marine debris that automate monitoring ocean life and health. Radiant MLHub hosts two geographically diverse marine debris training datasets that practitioners in any sector can use to track this global pollution.

What’s next

Read the concept note to learn more about being part of this journey, and mark your calendars to join the two online consultation sessions on October 4 and 18 at 5:00 PM CEST. If you have any questions, contact us at p.golbach@endeva.org.

Endeva is leading the design of the ii2030 challenge with support from us, the German Federal Ministry for Economic Cooperation and Development (BMZ), digilab, and FAIR Forward.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Radiant Earth Foundation

Radiant Earth Foundation

A non-profit on a mission to make insights from Earth observations and machine learning more accessible for global development organizations and communities.